Modeling Nature’s Contributions to People (NCP)
The 14 NCP in this analysis (Extended Data Fig. 1) were chosen to span development and climate goals, and to be mappable with spatially explicit data representing the period 2000-2020. We use European Space Agency (ESA) 2015, for land cover, Landscan 2017 for population47 (these were the most current data available at the time we began our analysis). We focus on “nature’s contributions” to key benefits of interest (e.g., security in food, water, hazards, materials, culture), meaning we partition out the role of natural and semi-natural ecosystems in producing those benefits. For food security, we include the contributions of pollination to crop production, vegetation-mediated atmospheric moisture recycling to crop and livestock production (included as a global NCP), grassland fodder production to livestock production, and wild riverine and marine fisheries. For water security, we include the contributions of water quality regulation, via sediment retention and nutrient retention, but not water yield since the role of ecosystems in determining the quantity of water is minimal (other than by evapotranspiration which is already captured in the vegetation-mediated moisture recycling, and regulation of timing of flows which is captured in flood risk reduction). For security of protection from natural hazards, we include flood risk reduction and coastal risk reduction. For materials, we include timber production, fuelwood production, and access to nature (which could be used for gathering, and also links to culture). For cultural benefits we include coral reef tourism (as the only globally mapped form of marine-based tourism) and access to nature again (which in addition to gathering also captures recreation or other uses of nearby greenspace). Finally, for climate security we include total ecosystem carbon storage (as a global NCP). Below we briefly summarize the models used to map these local NCP (Extended Data Table 1) and global NCP (Extended Data Table 2), full information on each model is available in the SI Methods.
Local NCP
1) Nitrogen retention to regulate water quality for downstream populations is modeled using the InVEST48 Nutrient Delivery Ratio model, which is based on fertilizer application, precipitation, topography, and the retention capacity of vegetation, and has been previously used in global applications49. The people benefitting from nitrogen retention are those who would otherwise be exposed to nitrogen contamination in their drinking water. In this analysis, the number of people downstream were calculated for every pixel of habitat, to provide a sense of which habitat potentially benefits the most people. Ideally, to map realized nitrogen retention, we would be able to convert biophysical service production into a measure of change in well-being, whether monetary, in health terms, or otherwise. However, the state of the science and data available globally precludes this for most services, so our proxy was the number of people downstream who could potentially benefit from that retention. NCP for nitrogen retention is expressed as nitrogen retention on natural and semi-natural pixels multiplied by the number of people downstream of those pixels. (See SI Methods Section 1 for more detail.)
2) Sediment retention to regulate water quality for downstream populations is modeled by adapting the InVEST Sediment Delivery Ratio (SDR) model, which maps overland sediment generation and delivery to the stream using the Revised Universal Soil Loss Equation (RUSLE) and a conductivity index based on the upslope and downslope areas of each pixel. Ideally sediment retention would be delineated for reservoirs, irrigation canals, or other water delivery infrastructure that is most impacted by sedimentation, but lacking a comprehensive global dataset identifying all such infrastructure, we again use the proxy of number of people downstream (as described for nitrogen retention, above). NCP for sediment retention is expressed as sediment retention on natural and semi-natural pixels multiplied by the number of people downstream of those pixels. (See SI Methods Section 2.)
3) Crop pollination is modeled with a simplified version of InVEST, mapping the potential contribution of wild pollinators to nutrition production based on the sufficiency of habitat surrounding farmland and the pollination dependency of crops49. NCP for crop pollination is expressed in terms of the average equivalent number of people fed by pollination-dependent crops, attributed to nearby ecosystems based on the area of pollinator habitat within pollinator flight distance of crops. (See SI Methods Section 3.)
4) Fodder production for livestock is modeled using Version 3 of Co$ting Nature50. Supply of fodder is calculated as the livestock-accessible tonnes of dry matter productivity for the non-cropland cover fraction and demand is estimated by the head count of livestock in a grid cell multiplied by the average biomass intake requirements per animal. The NCP as fodder production for livestock is expressed in terms of an index (0-1), rescaled from the realized service which is reported as the smaller of the supply or demand (if consumption exceeds productivity, the gap is assumed to be met with feed). The best available global inputs for dry matter productivity, livestock headcount, cropland and land cover are used as input. (See SI Methods Section 4.)
5) Timber production includes commercial (e.g., for trade/export) and domestic (e.g., for local use) timber, modeled using Version 3 of Co$ting Nature as two spatially mutually exclusive layers, because they represent two different sets of beneficiaries. NCP for timber production is expressed as an index (0-1) based on forest productivity and accessibility for harvest. Total potential sustainable supply of timber is estimated from the best available global above-ground carbon stock map multiplied by fractional tree cover for rural areas only. The sustainable harvest is calculated as the reciprocal of the number of years taken to develop the stock at the annual sequestration rate, according to dry matter productivity data. Demand is calculated differently for commercial vs. domestic timber based on different assumptions of accessibility. Commercial accessibility is defined as within six hours’ travel time of a population center of >50K people and on slope gradients <70%. Domestic accessibility is defined as areas inaccessible for commercial harvest and harvest rates are based on per capita consumption multiplied by population within 10 km. (See SI Methods Section 5.)
6) Fuelwood production is calculated as a byproduct of the timber model from Version 3 of Co$ting Nature. NCP for fuelwood production, like timber, is represented as an index (0-1) based on forest productivity and accessibility for harvest, but in this case specifically by rural people. Fuelwood can overlap spatially with domestic and commercial timber use, given that domestic and commercial timber harvest will not consume all sustainably available woody biomass in all places, due to the slope gradient limit and/or in places where demand is less than supply, and fuelwood is often a by-product of timber harvest. (See SI Methods Section 6.)
7) Flood regulation is modeled using Version 2 of WaterWorld31. To map nature’s influence on flood risk reduction, we identify the upstream places where canopies, wetlands, and soils (green storage) retain and slowly release rainfall, to the benefit of downstream communities on floodplains. NCP for flood regulation is expressed as an index (0-1) based on “green” water storage multiplied by the number of people downstream on floodplains. (See SI Methods Section 7.)
8) Access to nature is used as a proxy for numerous direct and indirect benefits to people, such as recreation, hunting and gathering, aesthetics, mental and physical health, and other cultural values that depend on the ability of people to access nature. This proxy-NCP is expressed as the number of urban and rural51 people within 1 hour (or 6 hours, for sensitivity analysis) travel of natural and semi-natural habitat, taking the least-cost path (by foot, road, rail or boat) across a friction layer14 (See SI Methods Section 8.)
9) Riverine fish catch is based on spatial disaggregation of nationally reported catch for 2007-201416 and updated to include catch estimated by household consumption surveys in 32 countries with severe underreporting52. Catches from large lakes were excluded. To spatially disaggregate the global catch of 13.3 x106 tonnes within country borders, a multiple linear regression model of total fish catch in river basins compiled from the literature was fitted with three predictor variables: population density, river discharge, and percent wetland cover (n=40, R2adj=0.69). NCP for riverine fish harvest is represented as metric tonnes of fish caught per km2 of land area per year, spatially allocated to the locations of the harvest. (See SI Methods Section 9).
10) Marine fish catch is based on the Sea Around Us data to map fish catch for 2010-2014 within 30 min grid cells across the ocean53,54. NCP for marine fish harvest is represented as metric tonnes of fish caught per km2 per year, spatially disaggregated to the locations of the catch. (See SI Methods Section 10.)
11) Coral reef tourism is taken from the Mapping Ocean Wealth dataset30, which reports the NCP for coral reef-associated tourism as dollars of tourism expenditure (in deciles 1-10). National expenditure data are spatially distributed based on three independent sources: hotel rooms from the commercial Global Accommodation Reference Database (GARD), dive shops and dive sites from Diveboard, and user-generated photos from the image-sharing website, Flickr. (See SI Methods Section 11).
12) Coastal risk reduction is modeled with InVEST for terrestrial and coastal/off-shore habitats55–58, updating previous global modeling49 through the inclusion of new data and projecting the value back to the habitat. Coastal risk reduction depends on the physical exposure to coastal hazards (based on wind, waves, sea level rise, geomorphology, bathymetry). with and without natural habitat to attenuate storm surge, and the people exposed. NCP for coastal risk reduction is expressed as a unitless index of the coastal risk reduced by habitat multiplied by the number of people within the protective distance of the habitat. (See SI Methods Section 12.)
Global NCP
1) Vulnerable terrestrial ecosystem carbon storage is mapped as the above-ground and below-ground ecosystem carbon lost in a ‘typical’ disturbance event, rather than the total stock20. This includes terrestrial and coastal (mangrove, salt marsh, seagrass) ecosystem carbon pools (aboveground, belowground, and soils), based on what carbon is likely to be released if the ecosystem were converted. (See SI Methods Section 13.)
2) Atmospheric moisture recycling is the process of water arising from the surface of the earth as evaporation, flowing through the atmosphere as water vapor, and returning to the surface of the earth as precipitation. Sources of evaporation include canopy interception, soil interception, soil evaporation, vegetation transpiration, and open water evaporation. We employed an Eulerian moisture tracking model, WAM-2 layers21, to quantify the source of moisture, where it travels through the atmosphere, and where it falls out downwind. The NCP of moisture recycling, which is to say the moisture associated with intact vegetation, is expressed as the volume of water evapo-transpired that falls on all rainfed productive land (cropland, rangelands, and working forests59). (See SI Methods Section 14).
Attribution of value to natural assets
The first step in identifying critical natural assets is to attribute the magnitude of benefits and, where possible, the number of beneficiaries, to the ecosystems providing those benefits (e.g., attributing the value of pollination occurring on croplands to the nearby habitat supplying the pollinators, or the coastal risk reduced and number of people protected along the coastline to offshore as well as onshore habitats). We define natural assets as natural and semi-natural terrestrial ecosystems (including semi-natural lands like rangelands and production forests, but excluding cropland, urban areas, bare areas, and permanent snow and ice; Extended Data Table 3) and inland and marine waters. Model outputs for pollination and coastal risk reduction are mapped back to habitat based on the pollinator flight distance (SI Methods Section 3) and protective distance of coastal habitat (SI Methods Section 12), respectively. For sediment and nitrogen retention, the count of people downstream of each habitat pixel was summed according to a hydrologic flow accumulation (SI Methods Section 1) and for nature access, the count of people was calculated for each pixels by summing the population pixels within 1 hour travel time according to a friction surface (SI Methods Section 8). All other model outputs are coarser than ESA resolution and are masked to the LULC types defined as natural assets relevant to that NCP (e.g., only forests for timber but excluding forests for grazing; Extended Data Table 3).
Optimization of NCP
Using integer linear programming (prioritizr, SI Methods Section 22), we identify minimum areas required 1) within each country’s land borders and marine Exclusive Economic Zones (EEZs) for the local NCP and 2) within all global land area (excluding Antarctica) and all countries’ combined EEZ area for the global NCP, to reach target levels (ranging from 5% to 100%) of every NCP. This optimization selects for the highest values across all NCP, providing the most benefit and/or to the greatest number of people, but not accounting for complementarity or redundancies of adjacent pixels (i.e., not dynamically optimizing after each pixel’s selection). We define the land and marine area for the 90% target as “critical natural assets” because the remaining 10% of aggregate NCP value requires disproportionately more area to achieve. Land and marine borders are based on Flanders Marine Institute (2020; Extended Data Table 2), and overlapping claims are excluded from the national analyses. The 12 local NCP are optimized for each country, then aggregated globally, while the two global NCP are optimized globally. In addition to these two main optimizations, we assess the sensitivity to scale by optimizing the 12 local NCP globally (instead of by each country), both with and without the 2 global NCP, and by substituting different scales of beneficiaries mapping for people downstream and for access to nature (Extended Data Table 5). We also assess the sensitivity of the area and location of critical natural assets (the optimization solution for the 90% target) to different NCP combinations. These variations include optimizing for each NCP individually, and optimizing for all NCP but dropping each local NCP from the set of 12 to evaluate its effect on the overall optimization (Extended Data Table 5). We also examine the correspondence between NCP and the robustness of these different solutions, by calculating the percentage of area shared by different pairs of services (Extended Data Table 4) or the percentage of area shared by all solutions (SI Table 1). We summarize the land and ocean areas required by country in SI Table 1.
Number of people benefiting from critical natural assets
We map the areas benefiting from critical natural assets in order to calculate the number of direct beneficiaries of these assets, and to compare the number of beneficiaries to the number of people living on the lands comprising these critical natural assets. For this analysis we are only able to include NCP for which the flow of the benefit can be spatially delineated: downstream water quality regulation (sediment retention, nitrogen retention), flood mitigation, nature access, fuelwood provision, and coastal risk reduction. The benefiting areas of some of the material NCP that are traded (fish, timber, livestock, crops that are pollinated) or the location of people who buy those traded goods are not easily mapped, so people benefiting from these NCP are not included in this analysis of beneficiaries. However, people within an hour of critical natural assets may provide a surrogate for many of the material NCP that are locally consumed. For water quality regulation, we take the population within the areas downstream (SI Methods Section 1) of critical natural assets. For nature access, we take the population within an hour’s travel (by foot, car, boat or rail; SI Methods Section 8) of critical natural assets. Likewise, for the other NCP we take the relevant population downstream, within the protective distance, or a gathering distance of critical natural assets. The relevant population for each NCP is considered to be the total global population for nature access and water quality regulation, but is limited to the total population living within 10 km of floodplains for flood mitigation, population along coastlines (in exposed areas: <10m above mean sea level) for coastal risk reduction, and rural poor populations for fuelwood. Total “local” beneficiaries are calculated through the intersection of areas benefiting from different NCP, to avoid double-counting people in areas of overlap. We calculate the number of people and percent of relevant population benefiting globally for each NCP (Fig. 2b) and the total “local” beneficiaries globally (Fig. 2a) and by country (SI Table 2).
Overlap analysis
We evaluate how well local and global critical natural assets align spatially with each other, and with biodiversity (terrestrial vertebrate species Area of Habitat (AOH)22; SI Methods Section 15) and cultural diversity (proxied by the number of Indigenous and non-migrant languages23; SI Methods Section 16), to identify synergies between these different potential priorities. To examine the level of overlap between areas identified as critical for the 12 local NCP vs. the 2 global NCP, we calculate the area (globally and by country; SI Table 1) where local NCP are selected by the optimization (at the 90% target level) and global NCP are not, where global NCP are selected by the optimization and local NCP are not, and where both are selected by their respective optimizations (the overlap). To calculate the species and languages represented by critical natural assets, we count the number of species whose AOH area targets overlaps these areas (SI Table 3) and the number of languages partially intersecting these areas (SI Table 4) globally and within each country. (See SI Methods Section 23 for more detail.)